Caffeine has been used as a metabolic probe to determine the relative levels of CYP1A2 activity in different individuals, since this compound is specifically 3-demethylated by CYP1A2. Urine specimen obtained at a 4-5-h interval after caffeine ingestion from 205 Japanese were analyzed using the [1,7-dimethyluric acid + 1,7-dimethylxanthine]/caffeine (1,3,7-trimethylxanthine) ratio, which better correlated with the rate constant for caffeine 3-demethylation than other ratios. The probit analyses of nonsmokers (n = 147) and smokers (n = 58) suggested that the CYP1A2 activity was not normally distributed and appeared bimodal. The breakpoints were at 5.0 and 6.0 of (1,7-dimethyluric acid + 1,7-dimethylxanthine)/1,3,7-trimethylxanthine ratio in nonsmokers and smokers, respectively. The bimodal probit plot suggested the existence of poor and extensive phenotypes. The percentage of individuals with the poor phenotype in Japanese was 14.1%. Induction of CYP1A2 by cigarette smoking was confirmed by the higher molar ratio observed in smokers (P < 0.0001). The CYP1A2 ratio was also higher in males than in females (P = 0.04). A reproducibility study of 12 subjects in an 11 month interval period showed that intraindividual variability did not alter this CYP1A2 phenotypic classification. Family study in eight pedigrees suggested that the poor phenotype of CYP1A2 inherited as an autosomal recessive trait. The sequences of CYP1A2 gene from poor and extensive metabolizers were analyzed. Although no differences of nucleotide sequence were observed in exons, exon-intron junctions and 5'-flanking regions (up to -2.6 kilobases) of CYP1A2 gene between each phenotype, there were some sequences which differed from the previous reported data. This is the first report in which the CYP1A2 phenotype and a genetic polymorphism in the CYP1A2 gene were comparably investigated.